Allocate data shards to clients based on the method described in the FedAvg paper. The data is sorted by label and then split into `shard_num *

Calculate noise multiplier for given privacy parameters. Args: sample_rate: Sampling rate (batch_size / dataset_size)

Dynamically loads a class from a string representation. Args: class_string (str): The string representation of the class, including th

Classifies features using a linear discriminant analysis (LDA) approach with adjustable beta for local and global statistics. Args:

Performs Linear Discriminant Analysis (LDA) classification. This method classifies input data using LDA by computing the coefficients

Solves for beta values used in the model. This function computes the beta means and covariance based on the provided features and lab

Computes features and collect labels from the provided dataloader. Args: dataloader (DataLoader): A DataLoader containing the inp

Partition the dataset according to the Dirichlet distribution using a specified concentration parameter, `alpha`. Args: targets (

Evaluating server model. Args: model: Used model. Defaults to None, which will fallback to `self.model`. Returns:

Train VAE, generate shared data, distribute shared data.

Client model finetuning. This function will only be activated in `test()`

This function is for outputting model parameters that asked by `client_id`. Args: client_id (int): The ID of query client

Gets a list of client data indices. Load and return the client-side data index from the partition file for the specified dataset. Ra

Load the specified dataset according to the configuration. Returns: BaseDataset: This is the loaded dataset instance, which i

Define data preprocessing schemes. These schemes will work for every client. Consider to overwrite this function for your unique data

Get DP configuration to send to clients with calculated sample_rate and noise_multiplier.

Partitions the dataset into IID (Independent and Identically Distributed) segments for each client. Args: targets (np.ndarray): A

Increment round counter after each training round.

Initialize data loaders for training, validation, and testing. Args: dataset: The dataset to be used for creating subsets. data_i

This function is for loading data indices for No.`self.client_id` client.

Package data that client needs to transmit to the server. You can override this function and add more parameters. Returns:

Calculate pairwise KL divergence between two sets of distributions.

Dynamically loads a partitioner class from a string representation of a flwr.datasets.Dataset. Args: flower_partitioner_class (st

Prepare and convert model for DP training. Returns DP-compatible model.

Partition data to ensure each client has a nearly equal distribution of classes. Args: targets (np.ndarray): Array of data labels

Maps hexadecimal class value (string) to a decimal number. returns: - 0 through 9 for classes representing respective numbers - 10 throug

The entrypoint of FL-bench experiment.

Save the learning curves of FL-bench experiment.

Save the metrics stats of FL-bench experiment.

Partition the dataset semantically using embeddings from a trained EfficientNet. Args: dataset (Dataset): The input dataset.

Set the LDA weights for the model classifier. This function computes the coefficients and intercepts for the linear discriminant analysis (LD

Inject server-side logger and output initialization message.

Show the maximum stats that FL method get.

The function for testing FL method's output (a single global model or personalized client models).

The function for testing FL method's output (a single global model or personalized client models).